Hackers Used Anthropic's Claude to Break into OpenAI

No paywall: https://www.theguardian.com/technology/2026/sep/18/openai-ha...

They also used Sol. It looks very bad. Either this is another stunt where both sides exaggerate the hack for public relations, or all nation states already have all the models.

I would believe that companies full of researchers who think the Python scientific stack is an example for good engineering have horrible practices.